Estimation of Acoustic Impedance for Surfaces Delimiting the Volume

of an Enclosed Space
Janusz PIECHOWICZ, Ireneusz CZAJKA
AGH University of Science and Technology
Al. A. Mickiewicza 30, 30-059 Krakow, Poland; e-mail: piechowi@agh.edu.pl
(received October 3, 2011; accepted February 17, 2012 )

Several methods can be applied for analyses of the acoustic field in enclosed rooms namely: wave
propagation, geometrical or statistical analysis. The paper presents problems related to application
of the boundary elements method to modelling of acoustic field parameters. Experimental and
numerical studies have been combined for evaluation of acoustic impedance of the material used
for the walls of a model room. The experimental studies have been carried out by implementing a
multichannel measuring system inside the constructed model of an industrial room. The measuring
system allowed simultaneous measurements of the source parameters – the loudspeaker membrane
vibration speed, the acoustic pressure values in reception points located inside the model space as
well as phase shifts between signals registered in various reception points. The numerical
modelling making use of the acoustic pressure values measured inside the analyzed space allowed
determination of requested parameters of the surface at the space boundary.
Keywords: acoustic impedance, acoustic measurement, numerical methods, sound field.

1. Introduction
Modelling of the distribution of acoustic field parameters in enclosed rooms is often
accompanied by a presence of considerable errors. Origins of those errors most often can be traced
down to phenomena taking place at the boundary of the analyzed volume (Kosała, 2008). Two
types of approach are encountered in description of those phenomena. A practical or engineering
approach is based on a description of the boundary defining its absorption coefficient. The
material’s compressibility or the surface structure de- tails are simply neglected. Such an approach
is widely applied because it offers a relatively easy procedure for determination of the absorption
coefficient values, both reverberation and physical ones. The problem that has to be coped with is
the necessity to mea- sure the values for all possible versions of the boundary elements fastening,
as it can have a considerable effect on the values of the absorption coefficients. In the second
approach that is applied in the acoustic field modelling by a wave propagation analysis the
phenomena taking place on the boundary are described by the values of the complex acoustic
impedance defining the dependence between the acoustic pressure and the acoustic particle
velocity at the boundary of the analyzed volume (Weyna, 2007). Such a description introduces
more serious practical problems. The task of determination of acoustic impedance values for real
model structures becomes much more difficult (Alba, 2011). There are methods offering the
solution of this problem but, similarly to the case of absorption coefficient determination,
modification of the boundary elements fastening may considerably change acoustic impedance
values of the surface. Because of the above, the authors have decided to combine the experimental
methods (Piechowicz, 2007a; 2007b) with numerica modelling of the acoustic field (Bj¨ork, 1987;
Kin- caid, 2005), which may lead to a successful evaluation of acoustic impedance for the surfaces
delimit- ing the analyzed volume with the enclosed acoustic field.
The analyzed problem is formulated as a task to determine such values of acoustical
impedance for surfaces delimiting the analyzed space that during the numerical modelling process
reproduce as closely as possible the measured results. The surface acoustical impedance values are
treated as optimal when an objective function built of differences between the experimental and
model values reaches the minimum. In order to solve such a problem it is necessary to execute the
following partial tasks:
• build a model connecting the impedance val- ues at the boundary of an analyzed volume
and the acoustic pressure values in selected reception points,
• define a function (norm) for determination of the interval between experimental and
simulation results,
• implement the model and the appropriate objective function.
2. Problems of acoustic field modelling in an enclosed space
Acoustic field in a selected volume of space can be modelled in transitional states (the
wave equation) or in a steady state (the Helmholtz equation) (Kuttruff, 1991). The process of
solving the differential equations used for modelling of the acoustic field can be practically realized
in many ways, among which the most frequently used are the ones listed below:
• analytical method – for simple room geometries and non-complicated boundary
conditions;
• finite element method – fast, universal method which fails for large volumes and higher
frequencies;
• boundary element methods – reduces the model rank by one, reducing 3D problems to
2D – which is, however, very demanding with respect to computational resources;
• non-grid methods.
In our work, the boundary element method has been used for modelling the acoustic field
in an enclosed room (Ciskowski, 1991; Gołaś, 1995; Kirkup, 1998). If the acoustic pressure is
labelled as p , the acoustic wave equation, with additional assumption that the solution is harmonic
in time, takes the following form (the Helmholtz equation):

where k is the wave number, it defines the relationship between the frequency ω and the velocity
of propagation c of the sound wave: k = ω/c. The boundary conditions imposed on the boundary
surface for this equation can be written as follows:
• Dirichlet condition:

• Neumann condition:

• impedance condition, called also mixed or Robin’s condition:

where p is the acoustic pressure, vn is the normal component of the particle velocity, ω is the

angular frequency, is the partial derivative with respect to the normal, ρ0 is the density of the
air ≈1.21 kg/m3, and Z is the acoustic impedance of a material. After employing the Green’s
identity the following integral boundary equation can be written in the space delimited by the
surface S:

where provides the fundamental solution, and cp is a coefficient dependent on

the point’s location.
After carrying out the boundary discretization and assuming appropriate shape functions
one obtains the following equation for every sub-surface (Ciskowski, 1991; Kirkup, 1998):

Each of the sub-surfaces is a boundary element. The behaviour of the variables in each element is
defined by a suitable shape function. The shape function can be constant, linear, or parabolic.
Constant shape functions were used by the authors, which means that in the whole sub surface
there is a constant acoustic pres- sure and velocity. This treatment makes determination of integral
values in the above equation much easier (Ciskowski, 1991). The above equation describes the
acoustic pressure in i-th node caused by the pressures and velocities in all other nodes. After
executing the following substitutions:

one can write:

Taking into the account that , where v is the acoustic velocity, one can execute a transformation
of the above formula. After performing a multiplication of the coefficients gij by iωρ0, the set of
equations can be expressed in a matrix form as:
Hp = Gv, (6)
where p, v are column vectors containing the node values for acoustic pressures and velocities
respectively, while H and G are squarematrices of coefficients called influence matrices.
A necessary condition for solving the problem is the knowledge of the acoustic pressure or
acoustic velocity value in each of the boundary nodes. Then, by the power of the equation
mentioned above, one can determine the unknown values of the velocity or pressure at the
boundary. The acoustic impedance of the boundary Z is taken into the account as a mixed (Robin’s)
boundary condition in the nodes in the following way:

which in the matrix form can be written as,

v = Ep, (8)
where the E is a diagonal matrix that contains the known values of the inverse of acoustic
impedance, or admittance, in the node points. After taking into the account the dependence, the
written above Eq. (6) takes the following form:
Hp = GEp. (9)
In the boundary element method the calculations are executed in two stages:
• determination of the acoustic pressure and velocity values in all boundary nodes,
• determination of the pressure values in the analyzed space in selected points.
Therefore, it is difficult to use directly Eq. (9) for determination of the acoustic impedance of the
delimiting surfaces. Equation (9) is just a starting equation for determining of the acoustic
impedance of the surfaces delimiting the room (Fig. 1).
The same acoustic impedance determination was carried out using the numerical model.
To test the room, a model was prepared in a 1:3 scale made of 6 Oriented Strand Boards. In one
of the walls a speaker driven sinusoidal signal was placed. A laser vibrometer was used to make
non-contact velocity and phase mea- surements of the speaker’s membrane. The linear array of 24
microphones was moved down to the length of the cuboid in steps of 5 cm on the central plane.
Each of the microphones recorded the sound pressure and phase shifts of the signal. Figure 2 shows
the distribution of the ratio of the sound pressure and vibration velocity at the frequency of 100 Hz
in the model room. The measurement results collected in the real model of the room (see Fig. 2)
have been used as starting values for the tuning process of the numeric model. The model tuning
has been carried out taking into the account the values of the acoustic impedance on the bounding
walls.
 Fig. 1. Model of the analyzed room and distribution of
the reception points used for identification of the boundary
conditions: S is the sound source, RP are reception points
(numbered in the figure on the bottom).

Fig. 2. Distribution of the ratio of the sound pressure and

vibration velocity at the frequency f = 100 Hz.
3. Determination of the acoustic impedance of the model surface
Determination of the acoustic impedance values fo the bounding surfaces was based on a
measurement of acoustic pressure values in selected reception points located inside the model
room and a simultaneous measurement of the signal phase shifts with respect to the stimulation
signal (Piechowicz, 2011).
The scheme of the experiment was as follows:
1. Selection of the reception points used in the model tuning;
2. Estimation of effective values of the acoustic pressure, phase shifts between the pressure and
velocity of the stimulating membrane, as well as the effective value of the membrane’s velocity;
3. Assumption of the impedance values at the volume’s boundary;
4. Estimation of the acoustic pressure values at selected points by numerical calculations;
5. Specification of the objective function and calculation of its minimum.
In order to evaluate the quality of the chosen numeric model in relation to the measured
values it is necessary to assume the form of the respective function (see Fig. 3). At the present
stage of the study the error measure has been assumed as a sum of squares of differences. If one
assumes that pi is the acoustic pres sure value at the i-th reception point, then one can write it down
in the form:

where bpi is the value of the acoustic pressure determined by the numerical calculations. The
objective function defined above is a good tool for evaluation of the model quality. However, it
should be remembered that the obtained function minimum might not ensure a complete identity
of the experimental and simulation results (Fig. 4).

Fig. 3. Shape of the objective function.

The shape of the objective function shown in Fig. 3 was obtained by iterative calculations
for successive values of the acoustic impedance for the boundary of the analyzed space. The
solution was searched for the minimum objective function values (Fig. 4). Determination of the
acoustic impedance values at the boundary of the analyzed volume has been carried out by an
 Fig. 4. Shape of the objective function in the vicinity of the
global minimum.
appropriate optimization process. Values of the acoustic pressure pi = p(xi, yi, zi) generated in the
analyzed volume by a source located at its boundary have been determined. Momentary values of
the acoustic pressure in the reception points have been registered together with momentary values
of the membrane vibration velocity for the stimulating loudspeaker. The results of those
measurements have been used to determine the pressure amplitudes and the signal phase shifts
with respect to the stimulating signal (Table 1).
Table 1. Measured values of the sound pressure, phase
shifts, and membrane vibration velocity for stimulating the
loudspeaker for 18 measurements points.
 A numerical model consistent with the layout of the measurement system has been
constructed. Acoustic impedance values Zi have been set on the surfaces delimiting the analyzed
volume, with each value constant within the respective wall surface. Then values of the acoustic
pressures bpi have been determined in 18 reception points (see Fig. 1) attributed to respective
measurement points in the real system. The collected data allowed determination of an ob- jective
function in the form given by Eq. (10). In the next step, the Nelder-Mead optimization method has
been applied, as implemented in the statistical software package “R”.

4. Numerical model of the room

The presented problem of determination of acoustic impedance values for the surfaces
delimiting the analyzed room has been solved by applying numerical modelling. Such an approach
turned out to be necessary because analytical determination of the requested impedance value from
Eq. (5) was not possible. The numerical model comprised a spatial mapping of the interior
geometry for the analyzed volume. The principal difference between the model and the study of
the real object (Fig. 5) was the assumption of stiff delimiting surfaces.

Fig. 5. Differences between the experimental and numeri-

cally determined values for individual reception points.

The elaborated model has been used for determination of acoustic impedance values of the
walls. The stimulation in the form of the acoustic velocity at the volume’s boundary has been
applied on a surface equivalent to the membrane area of the stimulat- ing loudspeaker. For all the
remaining surfaces of the walls one value of the acoustic impedance has been assumed, and it was
later subject to optimization. For the calculations 18 reception points have been se- lected, located
along two perpendicular straight lines (see Fig. 1).
As a result of the calculations, the acoustic impedance value Z on the OSB surface of the
delimiting walls has been determined and presented in Table 2.
Table 2. Specific acoustic impedance Z of Oriented Strand
Board (22 mm thick panel) as a function of the frequency.
5. Summary
The study presents a possibility of appointing a normal wall acoustic impedance using an
inverse BEM approach from a set of sound pressures at different points measured in an interior
steady-state sound field. This is important because distribution of the sound pressure in the interior
shows the effect of wall coverings on the sound field. For the calculation a model with stiff
boundary surfaces has been used, while in the real system the surfaces have been realized as boards
fastened along their edges. It seems that this fact is the source of errors in determination of the
acoustic impedance values of the delimiting surfaces. The error is introduced by a presence of an
additional component of the wall impedance, resulting from its elasticity. Another consequence of
such a fastening of the bounding boards is an inhomogeneous distribution of the acoustic
impedance value within the whole bounding surface.
The calculations have shown that scaling of the objective function values may be very
helpful for improving a convergence of the optimization procedure as a slow convergence seems
to be a problem for some starting points. The analyses have shown that the objective function
exhibits only one local minimum in the analyzed variable range covering h−10 000, 10 000i×h−10
000i, 10 000ii. It seems that it is also a global minimum. This allows determination of acoustic
impedance values for the indicated frequencies. Subsequent studies were conducted for various
configurations of wall materials. Refined computational algorithms for the model space will allow
a transfer of the method to determine the acoustic impedance of actual walls of small factory
rooms. Further research directions that should be indicated include determination of the influence
of reception points distribution on a determined value of the acoustic impedance of walls,
estimation of the error bound for a determined impedance value resulting from the assumption of
stiff boundary surfaces delimiting an analyzed volume. It would also seem plausible to test other
methods of the acoustic field analysis to find a solution of the discussed problem, for example,
some algorithms from the non-grid methods family.

Acknowledgement
The study has been carried out under the Ministry of Science and Higher Education N N
504 342536 research project, realized at the Chair of Mechanics and Vibroacoustics, AGH-UST
Cracow.

References
1. Alba J., del Rey R., Ramis J., Arenas J.P. (2011), An inverse method to obtain porosity, fibre
diameter and density of fibrous sound absorbing materials, Archives of Acoustics, 36, 3, 561–
574.
2. Bj¨ork ˚A., Dahlquist G. (1974), Numerical methods,Prentice-Hall.
3. Ciskowski R.D., Brebbia C.A. (1991), Boundaryelement methods in acoustics, Elsevier,
London-NewYork.
4. Gołaś A. (1995), Computer methods in interior andenvironmental acoustics [in Polish], AGH,
Cracow.
5. Kincaid D., Cheney W. (2002), Numerical analysis,Mathematics of Scientific Computing, Third
Edition,The Wadsworth Group, Brooks/Cole.
6. Kirkup S. (1998), The boundary element method inacoustics (Integral Equation Methods in
Engineering),
Integrated Sound Software.
7. Kosała K. (2008), Global index of the acoustic qualityof sacral buildings at incomplete
information, Archivesof Acoustics, 33, 2, 165–183.
8. Kuttruff H. (1991), Room Acoustics, Elsevier Ap-plied Science, London – New York.
9. Piechowicz J. (2007a), Determination of acousticfields in industrial room, Archives of
Acoustics, 32, 2,313–319.
10. Piechowicz J. (2007b), Acoustic field in the mechan-ical workshop, Archives of Acoustics, 32,
4, 221–226.
11. Piechowicz J. (2011), Estimating surface acousticimpedance with the inverse method,
International Jour-nal of Occupational Safety and Ergonomics, 17, 3, 271–276.
12. Weyna S. (2007), Some comments about the existingtheory of sound with comparison to the
experimentalresearch of vector effects in real-life acoustic near fields,Archives of Acoustics, 32,
4, 859–870.